[sdiy] Re: FPGA 2-OP FM MIDI Polysynth

Scott Nordlund gsn10 at hotmail.com
Mon May 4 01:41:55 CEST 2009

>>If you used the DX-7 algorthm, you might be able to cram many more
>>operators and voices on there.
> Where in tarnation can I get the pertinent details for that?

I know certain desktop calculators (Wang in particular, like the LOCI-2) used the same log math to speed up/simplify many of the operations.

I have a feeling, however, that the DX7's strange/unpleasant (IMO) envelope attack (which I brought up on another mailing list) is also partly due to the anti-log lookup table.  As far as I can recall, my Casio VZ-10M does the same thing, which may indicate how much Casio were "influenced" by Yamaha's architecture.

I believe they also shared the oscillator structure between operators and just multiplexed a bunch of different phase registers, etc. to get a bunch of different operators out of the same logic.  This was/is probably pretty common- I know my Ensoniq EPS offers a polyphony-versus-sample rate tradeoff option, presumably for the same reason.

And of course the quarter-sine (or quarter-log(sin)) lookup table that was (apparently) abused in the DX11, etc. to get other waveforms.

I tend to think that duplicating the log(sin) architecture wouldn't have any advantage, since there are already fast multipliers, but the quarter sine and operator sharing might free up a lot of space, if you're not doing that already.

At any rate, it should be pretty obvious that Yamaha put a lot of thought into the development of the DX7.  It certainly paid off, too.  I'd be interested to see how many of these optimizations were implemented in the Synclavier's FM voices.  It obviously wasn't nearly as refined as the DX7, considering the amount of hardware it takes to achieve something much simpler (though sound quality is an entirely different argument).

Insert movie times and more without leaving Hotmail®.

More information about the Synth-diy mailing list